Neural network model software and hardware collaborative encryption method

The invention discloses a neural network model software and hardware collaborative encryption method. The encryption chip is adopted to store and read the secret key, the encryption and decryption speed of the model is increased, long encryption and decryption time consumption caused by large model parameter quantity is avoided, the user-defined authentication operator is preset in the model, the hardware fingerprint information of the running terminal equipment is obtained through the user-defined operator in the running process, and the user experience is improved. And performing secondary verification on the corresponding terminal equipment fingerprint information and the authentication server to prevent the operation behavior of the unauthorized terminal. A model is encrypted by generating a Salt value corresponding to a corresponding hardware fingerprint and a secret key together, and the secret key and the Salt value are stored in an authentication server or an encryption chip, so that a user can conven